Search Issue Tracker
Fixed
Fixed in 1.9.0
Votes
48
Found in [Package]
1.8.1
Issue ID
ISXB-447
Regression
No
[Input System] Enabling both optimization options of Input System reduces performance in built Players
Steps to reproduce:
1. Open the attached user's project "RootingForYou.zip"
2. Open "ScoreManager.cs"
3. Enable and disable the performance optimizations used in the Awake() method
4. Make sure Development Build and Auto-connect Profiler is enabled
5. Once Built, open the Profiler and observe "PreUpdate.NewInputUpdate" Time ms
Expected results: Enabling both optimization options of Input System increases performance inbuilt Player
Actual Results: Enabling both optimization options of Input System reduces performance inbuilt Players
Reproducible: Input System 1.5.0 - 1.5.1 (2020.3.46f1, 2021.3.21f1, 2022.2.12f1, 2023.1.0b8)
Could not test on 2023.2.0a8 due to an Obsolete Custom package scripting errors
Reproducible on these devices:
VLNQA00415 - Oculus Quest 2 (Quest 2), CPU: Snapdragon XR2, GPU: Adreno 650, OS: 10
VLNQA00275 - Oculus Quest (Quest), CPU: Snapdragon 835 MSM8998, GPU: Adreno 540, OS: 10
N/A, Google Pixel 3 (Pixel 3), Android 12, CPU: Snapdragon 845, GPU: Adreno (TM) 630
VLNQA00413 - Samsung Galaxy Note10+ (SM-N975F), OS: 9, CPU: Exynos 9 Series 9825, GPU: Mali-G76
VLNQA00372 - Galaxy S21 5G (SM-G991U), CPU: Snapdragon 888, GPU: Adreno 660, OS: 11
VLNQA00202 - Samsung Galaxy Note 9 USA (SM-N9600), CPU: Snapdragon 845 SDM845, GPU: Adreno 630, OS: 9.0.0
VLNQA00177 - Xiaomi Mi Note 2 (Mi Note 2), OS: 6.0.1, CPU: Snapdragon 820 MSM8996, GPU: Adreno 530
Customer Devices:
Pico 4
Environment: Windows 10 22H2
Notes:
-Optimizations in question:
InputSystem.settings.SetInternalFeatureFlag("USE_OPTIMIZED_CONTROLS", true);
InputSystem.settings.SetInternalFeatureFlag("USE_READ_VALUE_CACHING", true);
-Issue is not exclusive to VR
-When Optimizations are enabled, Time in ms increases to 15-20ms from 5-9ms
-Issue does not reproduce on a non-VR Windows standalone Build
-
Liam-C
Sep 16, 2023 16:52
Any update on this? This bug report is almost 6 months old now, it has been under consideration for a long time, and there's been 6 input system updates since that don't address this. My quest game spends 1.3ms per frame on input processing on the main thread, so i'd love to be able to optimize that. However, I get the same 50-100% increase in CPU cost when enabling these flags (unity 2021.3.28)
Add comment
All about bugs
View bugs we have successfully reproduced, and vote for the bugs you want to see fixed most urgently.
Latest issues
- Asset Bundles retain their previous hash and CRC values when an object within a bundle is changed and rebuilt
- APV Reflection Probe Normalization breaks when SSGI is enabled
- Default Custom Components in project have Library counterparts
- [iOS]"The destination host has an erroneous SSL certificate" error is thrown when using UnityWebRequest to connect to the server with a self-signed certificate
- Freeze/crash on DynamicHeapAllocator::Allocate when opening a specific project
Resolution Note (fix version 1.9.0):
The issue was partially fixed, as only the `USE_OPTIMIZED_CONTROLS` flag was improved.
The Control Value Caching optimization path was not improved as it is by design, unfortunately. Only the documentation was updated so that users have more realistic expectations of its impact.
Fix will be available in upcoming Input System version 1.8.3